How to Shaping , Sizing Virtual resources , safe running VMs on VMware infra?.. And How to calculate how many VMs are running on single host server, it helps to face unplanned downtime when one host goes fail down??
-
Dear Team,
My one of the customer using vSphere Environments. They have 6 hosts in VMware cluster. They are running "88" Virtual machines.
I analyzed the Virtual infra, they are allocating vCPUs more than physical resources and also I have doubt, if one VMware host goes to downstate means,
what will be happening on VMware cluster??... I will try to analyze virtual hardware resources usage and all information I noted down here.
kindly guide me, how to shaping, sizing resources capacity and safe running Virtual infra in unplanned downtimes. Provide your suggestion.single host hardware resources :
Logical Processors: 32 [ with enabling Hyperthreading ]
CPU speed : 3.20 GHz [ 16 Cores * 3.20Ghz is 51.2 GHz].
Memory : 192 GBTotal No. of the host is "6" with same hardware resources on VMware cluster.
Total VMware cluster Resources in "6" host server:
Total vCPU is 32 * 6 : 192
Total CPU capacity : 306.82 GHz [ 51.2 * 6 * 3.20 GHz ]
Total Memory Capacity : 192GB * 6 : 1.12 TB
Total No. of Power ON VMs running is "88" .Right now, Total CPUs and Memory resources allocated on running Virtual Machines is
vCPU = 380 vCPU [ Total logical processors is 32*6 : 192 vCPUs]
Memory = 943596GB [ 921.480 TB ]
Thanks
Ghani -
There is no way in hell that your client has 921 TB of RAM allocated in this cluster. Something is wrong with what you've presented us.
In terms of CPU and RAM over allocation is completely fine, vSphere will actually tell you how the system as a whole is performing.
If the VM's are slow or sluggish then investigate things and see if the issue is a RAM issue, or a CPU or Disk IOPS issue.
In terms of the 6 hosts and worrying about a single host going down. Assuming they have the resources appropriately configured for this, then only 1/6th of the total resources would be unavailable during this outage.
-
@dustinb3403 said in How to Shaping , Sizing Virtual resources , safe running VMs on VMware infra?.. And How to calculate how many VMs are running on single host server, it helps to face unplanned downtime when one host goes fail down??:
vSphere will actually tell you how the system as a
how to calculate the number of VMs running each VMware host server? if it goes down one or two VMware host server. Based on current CPU & Memory allocation? ...
-
@ghani said in How to Shaping , Sizing Virtual resources , safe running VMs on VMware infra?.. And How to calculate how many VMs are running on single host server, it helps to face unplanned downtime when one host goes fail down??:
@dustinb3403 said in How to Shaping , Sizing Virtual resources , safe running VMs on VMware infra?.. And How to calculate how many VMs are running on single host server, it helps to face unplanned downtime when one host goes fail down??:
vSphere will actually tell you how the system as a
how to calculate the number of VMs running each VMware host server? if it goes down one or two VMware host server. Based on current CPU & Memory allocation? ...
This doesn't make any sense, please try explaining again. The servers could run just a single VM or they could run over 9000 VM's. It's all based on what the host has for CPU, RAM, IOPS and Storage, and then it depends on what each VM is configured with.
-
@ghani said in How to Shaping , Sizing Virtual resources , safe running VMs on VMware infra?.. And How to calculate how many VMs are running on single host server, it helps to face unplanned downtime when one host goes fail down??:
Total vCPU is 32 * 6 : 192
Total CPU capacity : 306.82 GHz [ 51.2 * 6 * 3.20 GHz ]
Total Memory Capacity : 192GB * 6 : 1.12 TB
Total No. of Power ON VMs running is "88" .
Right now, Total CPUs and Memory resources allocated on running Virtual Machines is
vCPU = 380 vCPU [ Total logical processors is 32*6 : 192 vCPUs]
Memory = 943596GB [ 921.480 TB ]So you have 1.12 TB of Physical RAM across 6 hosts, but have 88 VM's overallocated to 921 TB of RAM? You do realize that is like a 900:1 overallocation ratio, right? I don't install virtual environments often at all, but the best I've possibly heard of overallocation of any resource is 2:1, maybe 5:1 if you have it all right. 900:1 shows that something is not right with your calculations somewhere. That would also mean that you have on average 10.5 TB of RAM allocated to each and every VM in the cluster. Why would a VM need 10.5 TB of RAM? The most I have allocated to a VM is 32 GB of RAM for my SQL servers. Everything else is 16's and 8's. They don't need more than that.
As far as the environment itself, make sure that all of the hosts are in the same cluster and that DRS is enabled so that, in case a host does go down, all of the VM's will automatically be migrated to the rest of the online hosts until the downed host could be brought back online. DRS will ensure that the load between all of the hosts are balanced in vCPU and RAM load.
Also, is this an IPOD setup?
-
oh sorry , this is my mistake .. it is 921 GB .
-
they have standard license , so drs is not support features on that vmware environment
-
sorry .. what is IPOD setup / i dont hear about it //
-
@ghani said in How to Shaping , Sizing Virtual resources , safe running VMs on VMware infra?.. And How to calculate how many VMs are running on single host server, it helps to face unplanned downtime when one host goes fail down??:
sorry .. what is IPOD setup / i dont hear about it //
Inverted Pyramid of Doom is when you have a SAN (Storage Area Network) that stores all of the data for the VMs below (either directly or indirectly, say through dedicated switches) hosts. Number of hosts really doesn't matter as long as it is more than 1. If one host was to die, no problem, you can reboot that server with the latest snapshot in the SAN. If the SAN dies for some reason, then, well, you're out of luck because all of your VMs are all in the same storage platform.
Typically, hyperconvergence is more recommended because everything is setup in nodes, where each host has its own networking, compute, and storage components and the whole load is shared between the number of nodes you have. Recommended to have 3 nodes, but 2 will work.
@ghani said in How to Shaping , Sizing Virtual resources , safe running VMs on VMware infra?.. And How to calculate how many VMs are running on single host server, it helps to face unplanned downtime when one host goes fail down??:
they have standard license , so drs is not support features on that vmware environment
That means that if a host was to go offline for some reason, the VMs that were supported on that host will not come back online until you manually tell the VMs to come back online, if they're available.
@ghani said in How to Shaping , Sizing Virtual resources , safe running VMs on VMware infra?.. And How to calculate how many VMs are running on single host server, it helps to face unplanned downtime when one host goes fail down??:
oh sorry , this is my mistake .. it is 921 GB .
Eh, mistakes happen. Nobody is perfect.
-
Here is a good reference link for you.
What is the Inverted Pyramid of Doom?
-
@dustinb3403 said in How to Shaping , Sizing Virtual resources , safe running VMs on VMware infra?.. And How to calculate how many VMs are running on single host server, it helps to face unplanned downtime when one host goes fail down??:
There is no way in hell that your client has 921 TB of RAM allocated in this cluster. Something is wrong with what you've presented us.
These people exist. When I ask them who they work for they dodge the question and if I ask for their name I get something like "bob". I've asked not to question it when they have a TAM/SE sitting next to them confirming they are not actually crazy, they just work for.... interesting employers who do strange things....
-
@ghani said in How to Shaping , Sizing Virtual resources , safe running VMs on VMware infra?.. And How to calculate how many VMs are running on single host server, it helps to face unplanned downtime when one host goes fail down??:
kindly guide me, how to shaping, sizing resources capacity and safe running Virtual infra in unplanned downtimes. Provide your suggestion.
Turn on Admission Control will let you reserve resources for the event of an HA failure.
-
@nerdydad said in How to Shaping , Sizing Virtual resources , safe running VMs on VMware infra?.. And How to calculate how many VMs are running on single host server, it helps to face unplanned downtime when one host goes fail down??:
900:1 shows that something is not right with your calculations somewhere. That would also mean that you have on average 10.5 TB of RAM allocated to each and every VM in the cluster. Why would a VM need 10.5 TB of RAM? The
Someone doing build testing of in-memory database scaling (Functional, not actual performance) and they are using the SWAP to SSD to redirect the memory SWAP to an Intel Octane device, or NVMe drives so it doesn't completely crash
